The Ghost in the Machine is Us

The Nature of Technology

“The economy is nothing more than the clever organization of technologies to

provide what we need.”

How Technology Creates Itself

In the opening pages of The Nature of Technology, W. Brian Arthur admits to a kind of intellectual haunting: as a young man trained in engineering, he could manipulate equations but could not say what technology is. Professors offered him a buffet of half-definitions, applied science, engineering practice, industrial processes, but none grasped the elusive essence of technology, the “technology-ness of technology.” Decades later, after pioneering work in economics and complexity theory, Arthur returned to the question with fresh eyes. His conclusion was radical in its simplicity: technology is self-creating, combinatorial, and evolutionary.

Arthur argues that technology is not a random accumulation of gadgets, nor merely the applied fallout of science. It is an autopoietic system: new technologies arise from combinations of old ones. The jet engine did not descend from the internal combustion engine in a Darwinian lineage; it was born from the recombination of compressors, turbines, and combustion chambers, all elements with prior technological lives. He calls this “combinatorial evolution.” Novelty emerges not from a void, but from rearrangements of existing parts around newly captured natural phenomena. Radar did not sprout from radio by mutation; it harnessed electromagnetic reflection for a fresh purpose. MRI was not the offspring of X-ray imaging but the repurposing of nuclear magnetic resonance from physics into medicine.

The same logic explains today’s Artificial Intelligence. Large Language Models are not lightning bolts of inspiration; they are recombinant entities. The transformer architecture of 2017 combines with GPU clusters, themselves the result of decades of advances in semiconductor design, and with vast data-harvesting techniques enabled by the internet. Together, these elements capture a new “phenomenon”: the statistical patterns in human language. LLMs do not understand in the human sense; they exploit this phenomenon, much as radar exploits reflection. Already we see recursion at work, AI writing code to improve AI, designing chips to train the next generation of models. Technology, as Arthur predicted, is bootstrapping itself upward.

Russian Dolls

How, for example, does a Large Language Models (ChatGPT, Gemini, Claude, Perplexity) trained on text, not equipped with mathematical algorithms. “Yet somehow, they can add numbers correctly "in its head". How does a system trained to predict the next word in a sequence learn to calculate, say, 36+59, without writing out each step?” Arthur states:

“Technology is no longer a machine with fixed architecture carrying out a fixed function. It is a system, a network of functionalities—a metabolism of things-executing-things—that can sense its environment and recongure its actions to execute appropriately.”

Arthur insists we must open the black box. Technologies are recursive all the way down, Russian dolls of ingenuity. A jet engine is made of turbines, turbines of blades, blades of alloys, alloys of metallurgical processes. Each layer captures a natural effect, fire, resonance, magnetism, and presses it into service. So too with AI: algorithms inside architectures, inside training routines, inside hardware. Each is itself a technology, nested within others, feeding forward into novel assemblages. Technology is recursive structure incarnate.

This dual logic, recursion and combination, makes Arthur’s thesis as consequential as Darwin’s for biology. The stock of technologies bootstraps itself: the more components we have, the more possibilities for recombination. As William Fielding Ogburn put it a century ago, “the streetcar could not have been invented in the last glacial period.” The same applies now: ChatGPT could not have been built in 1995, when the necessary computational and data-building blocks did not exist. The economy, Arthur reminds us, is not a container for technologies but their offspring. The Industrial Revolution was the blooming of steam, metallurgy, and organizational devices like the joint-stock company. Our current AI revolution is no different.

Midwives

Arthur takes aim at the myth of the lone inventor. Edison or Whittle are less gods than midwives. Technologies beget technologies; the genius lies in recognizing the recombination that others miss. In this sense, OpenAI’s breakthrough is not sui generis but the latest instance of technological parentage, the stitching together of old parts into something newly alive.

Yet Arthur is no celebrant. He is blunt about the unease that shadows technology. It creates our world but estranges us from the nature we trust. He likens this to tectonic plates grinding together: our deepest hopes rest on technology, our deepest trust on nature. The 21st century, he warns, will be defined by that collision. AI exemplifies this dissonance. We hope it will solve our problems, accelerate science, improve medicine; yet we recoil at its uncanny artificiality. We are attuned to nature’s voice, not to machines mimicking it. As with gene editing and neural prosthetics, AI moves us from using nature to intervening within it, and that transition may provoke our deepest anxieties.

The brilliance of Arthur’s book is to restore seriousness to technology. For too long, science has been the elder sibling, technology the overlooked adjunct. Arthur inverts the hierarchy: economies, societies, and even sciences are scaffolded by technological advance.

“The economy,” he writes, “is nothing more than the clever organization of technologies to provide what we need.”

To ignore technology’s inner logic is to misunderstand the genesis of modern life.

But there is a question Arthur leaves less examined. If technologies evolve through recombination and recursion, who chooses which combinations to pursue? AI exists not only because it was possible, but because billions in capital and research effort pushed it forward. Combinatorial logic explains the menu of possibilities, but human values and politics decide what is ordered. The direction of technological evolution is not neutral. It is steered.

Arthur’s closing question lingers with new urgency. As we watch AI design the next generation of AI, the issue is no longer abstract. If technology truly creates itself, what is our role? Perhaps only this: to decide, before the plates grind irreversibly, what ends we want this self-creating machinery to serve.

Arthur himself reminds us of another truth: “Technology builds itself organically from itself.” That organic growth has always been driven by human imagination as much as by technical logic. From fire to flight, from radio waves to recombinant DNA, each leap reflects not just recombination but human curiosity and will.

If technologies create themselves, they do so through us, through our stubborn ingenuity, our capacity to see in old parts something startlingly new. It is a cause not only for foreboding but for wonder. We are, as Arthur writes, “entranced by the magic of technology,” and perhaps rightly so. For even as machines evolve, so too does our inventiveness. The ghost in the machine is, in no small measure, still us.

Stay curious

Colin